Early Detection of Heart Failure Using Electronic Health Records

Ng, Kenney; Steinhubl, Steven R.; deFilippi, Christopher R.; Dey, Sanjoy; Stewart, Walter F.

doi:10.1161/circoutcomes.116.002797

Cited by 93 publications

(37 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Based on electronic health record data in the primary care setting, these models included information on demographics, vital signs, diagnoses, medications, laboratory values, Framingham heart failure signs and symptoms, hospitalizations, and imaging. Although taking a large number of features into account, the best‐performing model had a receiver operating characteristic area under the curve of about 0.80 19. In the present study, a simple model including only 4 features (of which leg bioimpedance was one) provided slightly higher discriminatory capacity for incident heart failure, which highlights the strong predictive value of the variables included.…”

Section: Discussionmentioning

confidence: 56%

“…Machine learning has been employed before in the heart failure setting, but with the primary goal to develop machine learning–based prediction models18, 19 in contrast to the present study, where we proceeded with standard Cox proportional hazards models after the top features had been identified. In a recent study, machine learning algorithms (random forests, gradient boosting machine, and support vector machine) were compared with standard regression models (logistic regression, and Poisson regression) to detect readmission of patients with established heart failure.…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Bioimpedance and New‐Onset Heart Failure: A Longitudinal Study of >500 000 Individuals From the General Population

Lindholm

Fukaya

Leeper

et al. 2018

JAHA

View full text Add to dashboard Cite

BackgroundHeart failure constitutes a high burden on patients and society, but although lifetime risk is high, it is difficult to predict without costly or invasive testing. We aimed to establish new risk factors of heart failure, which potentially could enable early diagnosis and preemptive treatment.Methods and ResultsWe applied machine learning in the UK Biobank in an agnostic search of risk factors for heart failure in 500 451 individuals, excluding individuals with prior heart failure. Novel factors were then subjected to several in‐depth analyses, including multivariable Cox models of incident heart failure, and assessment of discrimination and calibration. Machine learning confirmed many known and putative risk factors for heart failure and identified several novel candidates. Mean reticulocyte volume appeared as one novel factor and leg bioimpedance another, the latter appearing as the most important new marker. Leg bioimpedance was lower in those who developed heart failure during an up to 9.8‐year follow‐up. When adjusting for known heart failure risk factors, leg bioimpedance was inversely related to heart failure (hazard ratio [95% confidence interval], 0.60 [0.48–0.73] and 0.75 [0.59–0.94], in age‐ and sex‐adjusted and fully adjusted models, respectively, comparing the upper versus lower quartile). A model including leg bioimpedance, age, sex, and self‐reported history of myocardial infarction showed good discrimination for future heart failure hospitalization (Concordance index [C‐index]=0.82) and good calibration.ConclusionsLeg bioimpedance is inversely associated with heart failure incidence in the general population. A simple model of exclusively noninvasive measures, combining leg bioimpedance with history of myocardial infarction, age, and sex provides accurate predictive capacity.

show abstract

Section: Discussionmentioning

confidence: 56%

Section: Discussionmentioning

confidence: 99%

Bioimpedance and New‐Onset Heart Failure: A Longitudinal Study of >500 000 Individuals From the General Population

Lindholm

Fukaya

Leeper

et al. 2018

JAHA

View full text Add to dashboard Cite

show abstract

“…Prior work on machine learning for early detection of disease has shown that inpatient encounters have been strongly predictive of early mortality. [5] While the SEER database does not include such features, a study could be performed from electronic health records (EHRs) as a data source.…”

Section: Discussionmentioning

confidence: 99%

“…Meanwhile, machine learning has shown to have strong potential for early detection of clinical endpoints in applications such as disease prediction [5][6][7][8] , readmission prediction [9][10][11] , drug adverse event prediction [12] , among others.…”

Section: Introductionmentioning

confidence: 99%

Machine learning based predictive model of early mortality in stage III and IV prostate cancer

Chen

2020

Preprint

View full text Add to dashboard Cite

Prostate cancer remains the third highest cause of cancer-related deaths. Metastatic prostate cancer could yield poor prognosis, however there is limited work on predictive models for clinical decision support in stage III and IV prostate cancer. We developed a machine learning model for predicting early mortality in prostate cancer (survival less than 21 months after initial diagnosis). A cohort of 10,303 patients was extracted from the Surveillance, Epidemiology and End Results (SEER) program. Features were constructed in several domains including demographics, histology of primary tumor, and metastatic sites. Feature selection was performed followed by regularized logistic regression. The model was evaluated using 5-fold cross validation and achieved 75.2% accuracy with AUC 0.649. Of the 19 most predictive features, all of them were validated to be clinically meaningful for prediction of early mortality. Our study serves as a framework for prediction of early mortality in patients with stage II and stage IV prostate cancer, and can be generalized to predictive modeling problems for other relevant clinical endpoints. Future work should involve integration of other data sources such as electronic health record and genomic or metabolomic data.

show abstract

“…The optimal size of a training set to develop automated methods for detecting heart failure has been estimated to be 4000 patients. 109 …”

Section: Using Machine Learning and Decision Supportmentioning

confidence: 99%

A manifesto for cardiovascular imaging: addressing the human factor†

Fraser

2017

European Heart Journal - Cardiovascular Imaging

View full text Add to dashboard Cite

Our use of modern cardiovascular imaging tools has not kept pace with their technological development. Diagnostic errors are common but seldom investigated systematically. Rather than more impressive pictures, our main goal should be more precise tests of function which we select because their appropriate use has therapeutic implications which in turn have a beneficial impact on morbidity or mortality. We should practise analytical thinking, use checklists to avoid diagnostic pitfalls, and apply strategies that will reduce biases and avoid overdiagnosis. We should develop normative databases, so that we can apply diagnostic algorithms that take account of variations with age and risk factors and that allow us to calculate pre-test probability and report the post-test probability of disease. We should report the imprecision of a test, or its confidence limits, so that reference change values can be considered in daily clinical practice. We should develop decision support tools to improve the quality and interpretation of diagnostic imaging, so that we choose the single best test irrespective of modality. New imaging tools should be evaluated rigorously, so that their diagnostic performance is established before they are widely disseminated; this should be a shared responsibility of manufacturers with clinicians, leading to cost-effective implementation. Trials should evaluate diagnostic strategies against independent reference criteria. We should exploit advances in machine learning to analyse digital data sets and identify those features that best predict prognosis or responses to treatment. Addressing these human factors will reap benefit for patients, while technological advances continue unpredictably.

show abstract

Early Detection of Heart Failure Using Electronic Health Records

Cited by 93 publications

References 13 publications

Bioimpedance and New‐Onset Heart Failure: A Longitudinal Study of >500 000 Individuals From the General Population

Bioimpedance and New‐Onset Heart Failure: A Longitudinal Study of >500 000 Individuals From the General Population

Machine learning based predictive model of early mortality in stage III and IV prostate cancer

A manifesto for cardiovascular imaging: addressing the human factor†

Contact Info

Product

Resources

About